esp-to-lsm-model

This model is a fine-tuned version of Helsinki-NLP/opus-mt-es-es on a Spanish Mexican Sign Language (MSL) glosses dataset. It achieves the following results on the evaluation set:

• Loss: 0.4524
• Bleu: 68.8807

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1.5e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
3.6287	1.0	75	2.6073	14.3097
1.886	2.0	150	1.5408	44.9883
1.2239	3.0	225	1.1446	60.7215
1.0309	4.0	300	0.9445	49.9656
0.7936	5.0	375	0.8136	51.1677
0.6785	6.0	450	0.7128	38.5475
0.571	7.0	525	0.6493	49.2921
0.4767	8.0	600	0.5980	67.6139
0.4361	9.0	675	0.5642	74.1258
0.3873	10.0	750	0.5409	73.4943
0.3141	11.0	825	0.5166	56.0140
0.3238	12.0	900	0.4993	75.9506
0.3202	13.0	975	0.4861	76.3040
0.2779	14.0	1050	0.4757	52.8020
0.2384	15.0	1125	0.4648	67.2947
0.2698	16.0	1200	0.4632	52.5347
0.2495	17.0	1275	0.4568	69.9258
0.2258	18.0	1350	0.4550	69.4897
0.2174	19.0	1425	0.4535	67.6997
0.2184	20.0	1500	0.4524	68.8807

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1